Re: [S] Pooled odds ratios

Goran Brostrom (gb@stat.umu.se)
Sat, 4 Jul 1998 11:32:45 +0200 (CEST)


On Sat, 4 Jul 1998, Dr. L. Y. Hin wrote:

[Changed Subject line]
> Dear S-Plus users,
>
> Does anyone know any method to calculate a pooled odd-ratio for
> meta-analysis, when the only available information is the values of odd
> ratio or adjusted odd ratio together with the corresponding confidence
> intervals? To make the matter worse, the contingency tables for calculating
> the odd raio were not reported.

Lin,

one possibility would be to calculate a weighted average of your
observed odds ratios, with weights proportional to the inverse of
the squared lengths of the confidence intervals. This is based
on a normal approximation, and also supposes that the individual
confidence levels are equal.

An alternative would be to do the same thing for the log odds
ratios. Your choice should be based on how the given confidence
were calculated. I.e., were they calculated on the log scale
and then transformed, or were they calculated directly for
the odds ratios? You can tell by looking at the symmetry, or
lack of symmetry around the estimated odds ratios.

Goran
---------------------------------------------------------
Goran Brostrom
Department of Statistics tel: +46 90 786-5223
Umea University fax: +46 90 786-6614
S-90187 Umea, Sweden e-mail: gb@stat.umu.se
----------------------------------------------------------

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news